Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

2.52.22 Autolock on walk away doesn't work - security issue

This site may earn commission on affiliate links.
Software is always bigger and more complicated than you would think. Software for automobiles covers dozens of control modules that interact with each other over communication links. The total lines of code can be staggering.

EVs do not need engine or transmission controllers, but they do need a motor controller, battery controller, DC-DC converter, and on-board charger. That is two more than an ICE! Hybrids are even more complicated, with 100 controllers in a Chevy Volt, and over one billion lines of code, IIRC.

GSP
Interesting.
It seems like, from a user perspective then, Tesla gets the hard stuff right and slips on the noncritical parts like the media player.
 
It's not.


It is.

Given how rare this reported problem appears to be, try thinking of what the circumstances required to trigger it must be and then
think about how likely you'd be to consider those circumstances either in the design or testing phase.

But something like walkaway locks shouldn't have many inputs. Car in park, seat pressure, fob location, summon status... can't think of anything else. A well designed, modular system would isolate that functionality so that once it works, it doesn't break. Some of the code I work with has been untouched in almost 20 years, with some of it rooted in the late 1960's. It handles multiple devices and interfaces with multiple systems in near real time. Yet, it virtually never just breaks spontaneously even though other parts are still actively developed. So when I say it seems easy, it seems a lot easier than my day job!

However, as Dilbert's boss once said, anything I don't understand must be easy to do.

BTW, I bet the door lock problem is associated with the fob profile change. I suspect some input is being filtered incorrectly based on profile.
 
I would hope not, but that's all too common. I could easily see code forks for X versus S or HW1 versus HW2, and missing merges across those branches.
The same manifestation does not mean the same bug. the chances something as safety critical as a car OS is subject to "oo I wonder which code to start from" is really unlikely. Much more likely a different corner case occurred. I agree with the consensus that the testing feels like it's not what it should be sometimes - but really there are a ton more factors in walk-away than those listed above.

I like the consistently safe behavior of the car is the focus which occasionally misses the smaller things - AP swerves notwithstanding.
 
  • Like
Reactions: chillaban
the chances something as safety critical as a car OS is subject to "oo I wonder which code to start from" is really unlikely
I literally spent the entirety last week fixing problems due to exactly that issue in an electronic health record system... emergency room triage stuff in this case. The new release was started from the wrong code point. In this case, it all got caught in QA, but I could see how it could very easily slip through. Maybe that (and a few decades of facepalms) makes me cynical, but I find it serves me best to start with version control diffs and blames than to start considering corner cases.

Anyway, not here to argue. Just bringing my perspective, for whatever it's worth.

I like the consistently safe behavior of the car is the focus which occasionally misses the smaller things - AP swerves notwithstanding.
100% agreement. In my safe place universe, the non-safety stuff (e.g., USB) is done by interns who are learning, and I give that a pass. :)
 
  • Like
Reactions: WannabeOwner
Hopefully Tesla has some regression tests so after they push software out they can compare stats from the cars in the fleet to see if something weird is going on (e.g., "now instead of 87% of the cars being locked, only 65% are", or "the average Wh/mi seems to have changed by more than 5%", or "seems like airbags are deploying more often since we changed the media player").
 
Hopefully Tesla has some regression tests so after they push software out they can compare stats from the cars in the fleet to see if something weird is going on (e.g., "now instead of 87% of the cars being locked, only 65% are", or "the average Wh/mi seems to have changed by more than 5%", or "seems like airbags are deploying more often since we changed the media player").

If they are logging drivers overriding of the wiper rain-sensing they certainly aren't getting around to fixing the algorithm!
 
  • Like
Reactions: Dithermaster